Web Page Fetch & Extract

Turn any live web page into clean, LLM-ready data — JavaScript fully rendered in a real browser, nav/ads/boilerplate stripped. Paste one or more URLs and get back page content as Markdown, plain text, or HTML, elements matched by a CSS selector, or a full-page screenshot. No scraping stack, no headless-browser plumbing, no proxy rotation to manage.

What it does

For each URL you pass, pick one operation:

• Fetch — the fully-rendered page as clean Markdown (great for RAG and LLM prompts), plain text, or cleaned HTML. Handles SPAs and JS-heavy sites that a plain HTTP GET returns empty or garbled.
• Extract — pass a CSS selector (h1, .price, table tr) and get back just the matching elements' text and attributes — perfect for prices, links, table rows, and headings.
• Screenshot — a pixel-accurate PNG (viewport or full-page, base64) rendered in a real browser, for visual verification or as input to a vision model.

Results land in the dataset, one row per URL — export as JSON, CSV, or Excel, or pull via the API.

Input

Field	Type	Description
urls	array	One or more absolute http(s) URLs. Required.
operation	string	fetch (default), extract, or screenshot.
format	string	markdown (default), text, or html. Applies to fetch.
selector	string	CSS selector. Required when operation = extract.
fullpage	boolean	Full-page vs viewport. Applies to screenshot. Default false.
width	integer	Viewport width in px (320–4096). Applies to screenshot. Default 1280.

Example input

{
  "urls": ["https://example.com"],
  "operation": "fetch",
  "format": "markdown"
}

Example output (one dataset row)

{
  "input_url": "https://example.com",
  "operation": "fetch",
  "http_status": 200,
  "format": "markdown",
  "content": "# Example Domain\n\nThis domain is for use in documentation examples..."
}

Use cases

• Feed clean page content into RAG pipelines and LLM summarizers
• Price / listing / table monitoring with a targeted CSS selector
• Visual verification and layout capture via screenshots
• Convert JS-rendered pages to Markdown for content archives and drafts

Pricing

Pay-per-result: you're charged once per URL successfully processed. No monthly minimum, no idle cost.

Notes

• URLs are validated and internal/private/file targets are blocked — the backend only fetches public web pages.
• A URL that fails returns a row with http_status and an error field instead of content, so one bad URL never fails the run.

Who runs this

Built and operated by CyberWareX — a suite of pay-per-call data and AI APIs for autonomous agents. Every result row carries provider / provider_url fields so the data's origin stays traceable wherever it travels. Full service directory and machine-readable docs: cyberwarex.com · daily correctness proofs: cyberwarex.com/trust.